Compare
As-of date financial news API
A date filter selects stories whose publication timestamp falls inside a window. An as-of cut selects the stories that were knowable at an instant. Those are the same set only while publication time and knowability time agree, and on SEC filings they often do not. Every product currently ranking for this phrase sells the first one.
Updated 2026-08-26. PIT figures read off a running server; vendor figures off each vendor’s own pages.
First
Which of the two you want
If the first row is your question, NewsAPI and Newscatcher will serve it better than we will.
| What you need | Buy | Because |
|---|---|---|
| “Give me stories about X published between two dates.” | NewsAPI.org, Newscatcher, or any general news feed with from and to. | They index a very large publisher set. We index SEC EDGAR and Federal Register public inspection, and nothing else. |
| “Give me the set a reader could have had on the evening of 9 March 2023, and tell me if any day in that window is missing from your ingest.” | PIT. | That answer needs a named clock, a null-safe rule for rows with no clock, and a per-day certificate. A publication-date filter has none of the three. |
Mechanism
Where a date filter and an as-of cut come apart
The two operations agree only when publication time and knowability time are the same number. On SEC filings they routinely are not.
EDGAR’s daily index carries a Date Filed and no time of day, so we stamp
the end of that day. published_at on the SVB receivership 8-K
reads 2023-03-10T23:59:59Z. The EDGAR receipt for the same
accession reads acceptance_at
2023-03-10T22:23:03Z. Stamping the start of the day instead
would have made that filing selectable at
2023-03-10T00:00:00Z, which is 19:00 Eastern on 9 March. A date
filter would then have handed you the whole of 10 March’s tape the previous
evening.
Measured across the corpus, 197 rows carry a Date Filed earlier than the index that first listed them. The tail runs long: one 2007 10-K first appears in the 2008-09-30 index, 581 days later. Filtering those rows on their filed date puts them in your window before anyone could read them.
Fit
Who each product is for
| NewsAPI.org | Newscatcher | PIT | |
|---|---|---|---|
| Job | REST search over a large publisher set, with top-headlines and date filters. The developer plan is localhost and non-production. | Structured news plus a CatchAll web search aimed at agents, with full text and NLP on the news API. | An as-of cut of SEC and Federal Register rows, identity resolved at the query instant, and a certificate per day touched. |
| Article body | Not on the listed plans; the URL is in the result and fetching it is between you and the publisher. | Yes, on the news API they advertise. | No body field exists on a row. You get form, accession, CIK, a constructed title, five clocks, a content hash and source_locator, which is the SEC URL. |
| Typical user | An app or dashboard that needs news about X this week. | A team that wants recall-first search and watchlists. | Someone whose backtest breaks if Friday’s filing shows up on Thursday. |
Clocks
What each product lets you filter on
| NewsAPI / Newscatcher | PIT | |
|---|---|---|
| Filter | from and to on article publication time, and on crawl time depending on vendor. | as_of with visible_by, inclusive and in UTC. A row whose named clock is null falls outside the cut instead of defaulting in. |
| Delay | NewsAPI developer runs 24 hours behind with a one-month archive; the paid rows are real-time with a five-year archive on Business and Advanced. | No delay product. A backfill is labeled by committed_at. Stamps taken at the time live in their own lane — a lane says how a row got its timestamp — and never mix with the certified one. |
| An empty answer | No matches. Whether the crawl covered that window is not something the response says. | HTTP 200 with count 0 when the day is certified; HTTP 409 with count as JSON null when it is not. |
| Google News and RSS | Those vendors index the open web. | Neither is certified here. We parse publisher bytes and keep the receipt for them. |
Corpus
What each side holds
PIT holds 587,689 queryable rows across seven sources, and counting. The SEC lane is 564,079 of them, with 204 days certified complete. Ten months of backfill are in. Five of those run day after day from 2022-11-01 to 2023-03-31, and the live tape has been writing new days since 2026-08-26.
| NewsAPI.org | Newscatcher | PIT | |
|---|---|---|---|
| Source count, as each vendor states it | 150,000-plus worldwide, 14 languages. | 140,000-plus sources, seven-plus years, custom sources on request. | Seven, and small on purpose: SEC EDGAR daily indexes across 18 form types, Federal Register public inspection, three FDA enforcement lanes, CFTC commitments, and OFAC. |
| YouTube and Reddit | Whatever reached a crawl. | Same class of problem if used as point-in-time. | Not ingested. The developer terms on both rule out building a point-in-time archive from them. |
| Identity | Keyword and source filters. | Entity resolution is a product they sell. | Ticker-to-CIK intervals valid at T, over nine tickers and sixteen intervals today. A join against the current ticker table is labeled leakage_risk=current_universe. |
Payload
The rows carry no filing text yet
The rows are metadata. A PIT row holds form type, company name, accession,
CIK, five clocks, a content hash and the SEC URL. The filing text is not in
there yet. A language model reading a cut therefore fills the detail in from
its own memory of the period, and that memory does not respect your cut
instant. Filing text is being ingested now. Until it lands, fetch the
document from source_locator and let the model read that
instead.
Price
Published list prices, August 2026
| Published band | |
|---|---|
| NewsAPI.org | Developer $0, development only, 100 requests a day, 24-hour delay, one-month archive. Business $449 a month for 250,000 requests and a five-year archive with no SLA. Advanced $1,749 a month for 2M requests and a 99.95% SLA. Enterprise is quoted, and extra requests bill on the paid rows. |
| Newscatcher | Individual $0, Starter $50, Scale $500 a month in credits, plus Enterprise. The news API runs a seven-day trial on the listed plans with production on Enterprise; CatchAll is per-record credits. |
| PIT | Researcher $29 and Power $79 a month, both reading the whole ingested history. There is no free API key; the watermarked playground is the free tier. Plans and what the corpus covers. |
Limits
Where NewsAPI is the better tool
- PIT indexes seven sources against their 150,000. For “any topic this morning”, that gap decides it.
- PIT has no full-text search and no NLP layer, so Newscatcher’s article bodies have no equivalent here.
- PIT ingests only sources whose terms allow a certified point-in-time archive, which rules out scraped publisher HTML.
- PIT’s certified backfill ends 2023-03-31 and the tape starts 2026-08-26, so a date filter aimed at 2024 or 2025 returns nothing.
- If the job really is “historical news with date filters”, the products above answer it and PIT does not.
Check it
The as-of cut, run twice
Same ticker, two instants twenty-eight hours apart. The 8-K is absent from the first and present in the second. The first answer states that both sources were certified complete for 9 March, so its zero means nobody filed.
No account needed. Responses carry sample: true.
Nearby